The Raspberry Pi 3 Compute Module Is On Its Way

The Raspberry Pi Foundation founder Eben Upton has revealed in an interview with PCWorld that there will be a new version of the organisation’s Compute Module featuring the faster processor from the latest Raspberry Pi 3 boards, and it will be available “In a few months”.

The Compute Module was always something of an odd one out among the Raspberry Pi range, being a stripped-out Raspberry Pi chipset on a SODIMM form factor card without peripherals for use as an embedded computer rather than the standalone card with all the interfaces we are used to in the other Pi boards. It has found a home as the unseen brains behind a selection of commercial products, and though there are a few interface boards for developers and experimenters available for it we haven’t seen a lot of it in the world of hackers and makers. Some have questioned its relevance when the outwardly similar Pi Zero can be had for a lower price, but this misses the point that the two boards have been created for completely different markets.

The Pi 3’s 1.2 GHz 64-bit quad-core ARM Cortex-A53 BCM2837 SoC will certainly up the ante in the Compute module’s market, but it will be interesting to see what changes if any they make to its form factor. The Foundation’s close ties with Broadcom mean that they have done an impressive job in maintaining backward compatibility at a hardware level between the different generations of their product, but it is unclear whether this extends to the possibility of the new module maintaining a pin-for-pin compatibility with the old. We’d expect this to be an unlikely prospect.

It is certain that we will see a new generation of exciting commercial products emerging based around the new module, but will we see it making waves within our domain? This will depend on its marketing, and in particular the price point and quantity purchase they set for it. The previous board when added to a Compute Module Development board was an expensive prospect compared to a Raspberry Pi Model B that became more unattractive still as newer Pi boards gained more capabilities. If they price this one competitively and perhaps if any cheaper open hardware breakout boards emerge for it, we could have a valuable new platform on our hands.

Here’s our coverage of the original Compute Module launch, back in 2014.

[via Liliputing and reddit].

BCM2837 image: By Jose.gil (Own work) [CC BY-SA 4.0], via Wikimedia Commons.

52 thoughts on “The Raspberry Pi 3 Compute Module Is On Its Way

        1. That’s a good idea!

          Using a DIMM socket means they can just use the PCB itself as the other connector. Nice and cheap, no unneccessary spending for someone who might want to buy a lot of these, and put them together on some sort of complicated PCB. If you’re doing a PCB like that, you likely have the capability of adding DIMM slots. If you don’t, then get a normal Pi. Since you’re obviously not a big business, the low quantities you’ll be working mean the extra money doesn’t make much difference.

          That said… Do these have a real advantage over Intel chips for computing servers? Is there a need for low wattage servers? That isn’t better served by Intel’s low-power chips? I thought Intel pretty much had every advantage if you want a lot of MIPS. Or AMD of course. What’s the use for this module, apart from hackers playing with it?

          There’s 8-core ARMs, and I think even more than that, but ARMs seem to be for low-wattage low-MIPS applications. Phones in particular. And phones only need so much CPU power, especially with the diminishing returns each extra core gets you. I think there’s a practical limit to how many ARM cores anyone could actually use.

          Also, why don’t ARMs run faster? Why are there no versions that do 2GHz or more?

          1. > Do these have a real advantage over Intel chips for computing servers?

            Wait wait wait, absolutely not! this is a SoC MCU, it’s meant to do IO with other chips on a PCB, you don’t want 5UART, 3 I2C, 4 SPI devices, ADCs and loads of GPIOs into a server CPU.

          2. I forgot to explain what’s the point of a System On a Module like the raspberry pi compute: you need a very fine pitch, high layer count high speed design around the MCU, because it has a fast clock and some external RAM. But when you design your system, you might not want to pay for high end designers and PCBs. So you buy a System On a Module, you design a low-speed, 2 layers cheap PCB, you put a DIMM connector on it and plug the SOM in it. That way the maker of the SOM benefits from the economy of scale, and you have less costs because you can use lesser technologies in your system.

      1. My local microcenter has had them in stock since the release (limit 1 per person per visit). Please revise your statement to “Comparing anything to the Pi Zero is kind of moot given that it is impossible *for me* to get *my* hands on one.’

        1. If I need only one unit, it doesn’t matter whether it costs 5USD or 20USD. If I need ten units, there is no difference between both options, if I have to pay additional 15USD for each single 5USD-unit in terms of shipping or in terms of gas+time and labor. If I could have 10 units next year at the earliest, I don’t care how much one single unit costs today; especially if I need 10 units today.

          Now, with a new compute module announced, there are still two options: Cheap or available. Two “completely different markets” indeed. Customer can choose.

          1. It must be nice to have money.
            Ive been trying in canada to get my hands on one that doent cost more then a PI 3, and then it is still cheaper to get a originel PI 3 from china, cause the shipping is free.

        2. Revised, on behalf of Tucson Tom:

          It’s impossible for 99% of the world to get their hands on them, as online retailers sell out of them in minutes and local retailers, if you’re lucky enough to have one, may not carry them at all, much less be able to keep them stocked.

          You’re the exception, mate. Stop thinking of yourself as the rule, yeah?

    1. You snooze, you loose… ordered mine the 2nd day they came out and had it 2 weeks later…
      Many places did offer them for quite some time on a 1 per person deal.

      1. The one per person deal is crap though.. Here in Europe shipping costs are more than what the zero costs. So it’s a huge waste if you can’t simply order 10 of them or so.

          1. I signed up here
            They run out of stock fairly quickly, but they keep being available longer. This one will just show adafruit stock. Last week I bought one each from pimoroni and thepihut, both in the UK, but they came out to ~$10 US with shipping each, the one from thepihut got here in less than a week. I just bought another because when looked up the link for the stock notifier it said they were available, I probably have too many actually.

  1. So, new compute module. Cool.
    I get it that they want to push this board. But out of the Raspberry pi lineup, this is probably the board with the less people caring about it… I’ve heard things about a Raspberry pi 3 A+, and THAT would be huge :)
    Unless the new compute module can be had with things to use it for a reasonable price, but even then, for most users, it’s of little use…

    1. But the commercial and industrial partners care. Some people still viciously deny that the competitive commercial market is a major priority for the Pi Foundation. But the compute module seems to suggest it is.

      1. The way I understand it through the contacts I’ve built up running my local Raspberry Jam is that the different parties involved (RS, Farnell & the Pi Foundation) in bringing the Pi to market are responsible for the different priorities in the products. The Compute modules I understand to be largely an RS effort, and their focus has always been into the industrial sector.

        1. For me it was an interesting product at my last job, I wanted a cheap computer on a module in small quantities (batches of 10-50), the zero with it’s very small connectors would have qualified, if I could source it reliably. I think consumer stuff is interesting when the volume is not too big, because you have help all over the internet, the docs are mostly not under NDA, and when it’s time to hire you don’t need to look for 15+ years experience veterans with they $10k+ tools. If your company grows it will be time to bring the expensive graybeards, nothing is lost.

    2. I want one that’s just the processor and RAM, power and mini HDMI. Maybe the GPIO around the edges. Not much bigger than the size of the processor so I can use it for an open source smart watch. And while I’m dreaming, let’s make it three dollars, impossible to get anywhere, and call it the Pi Nano. Or maybe Lambda, Idk. But I just got a small NTSC screen, and the Zero is twice the size of the screen! That’s ridiculous

  2. Interesting. How long will this module be available? For a new application we need it to be available for 10 years, for industrial even longer. The Raspberry Pi model B already seems near end of life… Yes, we can stockpile. But we don’t want to if we don’t have to.

    1. It’s not even worth using for industrial control. They were designed to be throw-a-way items used by students,

      If you need industrial control go to the people who specialize in it like Siemens, Rockwell, Allen-Bradley.

    2. “How long will this module be available?”
      Unless you are working one something that must be certified in someway who cares?
      As long as they keep the pinouts and form factor the same it shouldn’t be a problem. The pi is a Linux based system so your apps should work on any newer pi. The only issue could be device drivers and Kernels and that should be manageable.
      It is a lot better than rolling your own only to find some chip is EOL with no replacement available.

    3. You would be crazy to NOT consider iMX6, iMX6UL or iMX7 that have guarantee production dates for 10 years at least.

      96boards has common connectors and open source designs.

      Toradex has cheap ads sbcs

  3. Meh, by the time you build a interface board for this novelty item, it’ll cost more than a PI 3.

    Really at that size just a get a M4 ARM controller board for $10 and learn how to program on bare metal. You don’t need gigs of ram and flash for most embedded applications.

      1. The article is basically marketing blather for another Broadcom offering.

        And apparently judging from other posts no one has a clue either, except you and you won’t say.

    1. So if I wanted to make a stereo system that used GNU Radio and an rtl-sdr, and or a Si4777 for radio, bluetooth support for audio in and out, Then I want to add support for Pandora, Google Music, and maybe even voice control using sphinx and google speech recognition or Amazons speech recognition. Of course I want it to use Wifi to connect to my home network or maybe and ethernet cable and I want a front panel with hard controls. I might even want to have some audio switching capability as and the ability to read data from a USB drive and to play MP3s and FLAC off my NAS.
      So you really think an M4 would be a good choice?

  4. I really hope they can maintain compatibility. I have a CM in application, and being able to swap out CM3 would be very attractive. Having to retool the application, would not be attractive at all. If it’s not a drop-in, we lose any benefit of it being pluggable, and it’s over.

  5. One would assume that the notorious overheating problem exhibited by the Rpi3, which was ceremoniously swept under the (fireproof) rug by all concerned–and mostly unconcerned–Raspberyy Pi individuals has been fixed–but not reported as fixed, as that would constitute admission of the problem.
    Or maybe not, based on the history of how the original problem was not handled by RPi
    Don’t waste your time tracking down an answer, Hackaday Honchos. Raspberry Pi could give lessons in circumlocution to the candidates for POTUS.
    As a reference, all you Hackaday cognicenti, START here–

    and then quadruple your fun by typing “raspberry pi 3 overheating problem”, or/and “raspberry pi 3 heat problem halfacree” into your favorite search engine.


  6. “Can someone explain in simple terms…”
    First, and foremost, remember this fact: the Compute Module is NOT meant to be sold to you, as an individual user. It is meant to be sold in lots of one thousand or more to people who will design the device into a larger piece of equipment where a processing device is needed.
    Unfortunately for the RPi Foundation, the stated goals of the Foundation–teaching children to compute–are the same factors which severely cripple the Compute Module’s ability to be seriously considered as an embedded processor for high-speed, real time applications–there exists NO no real interrupt-processing capability, for example.
    Now you know why you don’t hear or see much of the Compute Module; nor will you ever, as long as the CM is based on the “teach-children-to-compute Raspberry Pi.

  7. I’m more interested in when the Pi2 & Pi3-based Model As are coming. I’d love me some of that.

    Pi Zeros are great (and yes, I do have some) but I’d love the extra grunt without having to power an onboard hub chipset / modding.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.